Background
Andrea Bartz, Charles Graeber and Kirk Wallace Johnson brought a copyright infringement suit against Anthropic PBC, alleging that their books were used to train Claude LLMs. The court granted summary judgment in part: use of legitimately purchased books for training was held to be fair use, but the retention and use of pirated copies was not, leaving liability and damages to be determined at trial.
AI interaction
The court held that training large language models on lawfully obtained works was “quintessentially transformative,” but stressed that “every factor points against fair use” where pirated copies were involved. This decision is the first major federal ruling to squarely apply fair use to AI training, clarifying that provenance of training data is legally decisive and shaping the compliance standards for future AI developers.